Intelligent security early warning method and system based on AIoT

Abstract

The invention relates to an AIoT-based intelligent security early warning method and system, and relates to the field of security, and the method comprises the steps: collecting illumination detection information and temperature parameters in a natural environment in real time; generating environment adjustment information based on the illumination detection information and the temperature parameters and outputting the environment adjustment information to acquisition equipment to acquire multi-modal data; generating behavior information based on the multi-modal data; when an abnormal behavior exists in the behavior information, generating early warning information based on the environment adjustment information and the behavior information; and triggering a hierarchical early warning mechanism based on the early warning information, and encrypting and transmitting the early warning information to a security system terminal for early warning. The method has the effect of reducing missed judgment or misjudgment of the security and protection system.

Description

Technical Field

[0001] The present invention relates to the field of security, and in particular to an intelligent security early warning method and system based on AIoT. Background Art

[0002] Security is a systematic project that uses comprehensive measures such as technical means, equipment deployment, and management systems to prevent, stop, and respond to various security threats, such as intrusion, theft, destruction, and disasters, to ensure the safety of personnel, property, information, and the environment.

[0003] At present, traditional security systems generally deploy cameras in dispersed locations that need to be monitored, and transmit the monitoring images to the management room through the cameras, relying on manual monitoring of the monitoring images in the management room.

[0004] Regarding the above-mentioned related technologies, traditional security systems rely on manual monitoring and are unable to concentrate on observation for a long time, resulting in missed judgments or misjudgments, which seriously affects the reliability of the security system. Summary of the Invention

[0005] In order to reduce missed or misjudgment of the security system, the present invention provides an intelligent security early warning method and system based on AIoT.

[0006] In the first aspect, the present invention provides an intelligent security early warning method based on AIoT, which adopts the following technical solutions: An intelligent security early warning method based on AIoT, comprising: S1: Real-time collection of light detection information and temperature parameters in the natural environment; S2: generating environmental adjustment information based on the light detection information and the temperature parameter and outputting the environmental adjustment information to a collection device to collect multimodal data; S3: generating behavioral information based on the multimodal data; S4: When the behavior information contains abnormal behavior, generating warning information based on the environment adjustment information and the behavior information; S5: triggering a hierarchical warning mechanism based on the warning information, and encrypting and transmitting the warning information to the security system terminal for warning.

[0007] By adopting the above technical solution, the acquisition equipment settings are dynamically optimized through real-time collection of light and temperature parameters to collect multimodal data and analyze it, thereby improving the accuracy and real-time performance of abnormal behavior identification in complex environments; at the same time, the acquired data is encrypted for transmission to improve the security of information transmission.

[0008] Optionally, generating the environment adjustment information based on the light detection information and the temperature parameter further includes: S21: When the light intensity in the light detection information is lower than a preset light reference intensity, obtaining a change rate of light shading, shape characteristics of the shading area, and a movement trajectory within a preset time period from the light detection information; S22: establishing a dynamic occlusion model based on the change rate, the shape feature, and the movement trajectory; S23: Determine whether the dynamic occlusion model conforms to a preset artificial dynamic occlusion model; S24: If the conditions are met, a preset drone is controlled based on the movement trajectory and the shape feature to perform airspace filling and collect modal filling data, and a preset collection device is controlled based on the movement trajectory to perform reverse displacement compensation, and the collected data is used as compensation data; S25: generating spatiotemporal correlation data based on the modal filling data and the compensation data; S26: generating spatiotemporal adjustment information based on the spatiotemporal correlation data and the temperature parameter, and using the spatiotemporal adjustment information as the environment adjustment information; S27: If not, generating obstacle avoidance information based on the dynamic occlusion model; S28: Generate environmental fusion information based on the obstacle avoidance information and the temperature parameter, and use the environmental fusion information as the environmental adjustment information.

[0009] Optionally, generating the environment adjustment information based on the light detection information and the temperature parameter further includes: S291: When the illumination intensity of the illumination detection information is higher than a preset illumination reference intensity, defining a region higher than the preset illumination reference intensity as a highlight region, and determining whether the temperature parameter is higher than a preset temperature reference parameter. S292: If the temperature is higher than a preset temperature reference parameter, define the area higher than the preset temperature reference parameter as a high temperature area; S293: generating a high-light-heat zone based on the high-light zone and the high-temperature zone; S294: Obtain the current preset position of the light shield as the current position of the baffle; S295: Calculating a rotation angle of the light shielding plate as an adjustment angle based on the current position of the baffle and the high-light hot zone; S296: Generate device adjustment information based on the adjustment angle; S297: Controlling the preset shading plate to execute based on the device adjustment information, and simultaneously obtaining illumination adjustment detection information and temperature adjustment parameters after execution; S298: Generate an environment control factor based on the light control detection information and the temperature control parameter, and use the environment control factor as the environment adjustment information; S299: If the temperature is not higher than the preset temperature reference parameter, retrieving the sound data from the multimodal data; S29A: Optimizing image acquisition parameters based on the illumination detection information; S29B: Generate audio and video adjustment information based on the image acquisition parameters and the sound data, and use the audio and video adjustment information as the environment adjustment information.

[0010] Optionally, generating device adjustment information based on the adjustment angle includes: S2961: Get monitoring area; S2962: When the monitoring area is inconsistent with the preset monitoring area, the unmonitored area is named as a monitoring blind area; S2963: Obtaining a coordinate range of the monitoring blind spot as the blind spot coordinates based on a preset spatial coordinate system; S2964: Calculating a monitoring area difference based on the blind spot coordinates; S2965: Calculating a rotation angle difference and a pitch angle difference based on the monitoring area difference and controlling the light intensity sensor to execute; S2966: Generate a focus parameter correction value based on the rotation angle difference, the pitch angle difference, and the blind spot coordinates; S2967: Generate the device adjustment information based on the focus parameter correction amount, the rotation angle difference, and the pitch angle difference.

[0011] Optional methods for determining abnormal behavior include: S61: extracting behavioral features from the multimodal data; S62: When the behavior feature does not conform to the preset behavior benchmark model, extracting items that do not conform to the behavior benchmark model from the behavior feature and marking them as abnormal items; S63: extracting the number of abnormal items and using it as the abnormal number; S64: generating anomaly confidence based on the abnormal item and the abnormal quantity; S65: Determine whether the abnormality confidence exceeds a preset abnormality reference confidence; S66: If it exceeds, it is determined that the abnormal behavior exists; S67: If not, the number of occurrences of the abnormal item is retrieved from the behavior feature as the number of abnormalities; S68: If the number of abnormalities exceeds a preset number, a secondary verification is performed based on the multimodal data.

[0012] Optionally, performing secondary verification based on the multimodal data includes: S681: Retrieving sound data from the multimodal data; S682: Generating voice behavior feature data based on the voice data and the behavior feature; S683: Matching the acoustic pattern feature data with a preset acoustic pattern database and extracting the one with the highest matching degree as an initial abnormality judgment value; S684: Calculate the difference between the initial abnormality judgment value and the reference matching degree as the abnormality difference; S685: When the abnormal difference exceeds a preset abnormal difference interval, it is determined that the abnormal behavior exists.

[0013] Optionally, the method for determining whether abnormal behavior exists also includes: S71: Obtaining current crowd density from the multimodal data; S72: Determine whether the current crowd density exceeds a preset density value; S73: If the density value is exceeded, the movement trajectory, movement speed and distribution characteristics of the crowd are collected in real time; S74: generating group dynamic data based on the movement trajectory, the movement speed, and the distribution characteristics; S75: Establishing a group dynamics model based on the group dynamics data and the current crowd density, and marking individual behavior characteristics; S76: generating sound behavior data based on the individual behavior characteristics and the sound data; S77: If the sound behavior data does not match the preset sound behavior reference data within the preset time period, the abnormal individual is marked and determined to have the abnormal behavior.

[0014] Optionally, establishing a group dynamics model based on the group dynamics data and the current crowd density includes: S751: generating an initial dynamic model based on the group dynamic data; S752: Determine whether the initial dynamic model is consistent with the group behavior benchmark model based on the initial dynamic model; S753: If they match, adjusting the initial dynamic model based on the current crowd density to form a group dynamic model; S754: If not, extracting an abnormal area from the initial dynamic model; S755: Calculating the area change rate of the abnormal region within a preset time period to generate the diffusion speed of the abnormal region; S756: Calculating a group risk index based on the diffusion speed of the abnormal area and the current crowd density; S757: Establishing a multi-layer group model based on the initial dynamic model, the abnormal area, and the risk index, and using the multi-layer group model as the group dynamic model.

[0015] Optionally, after encrypting and transmitting the warning information to the security system terminal for warning, the method further includes: S51: generating a modified risk index based on the environmental adjustment information and the multimodal data; S52: generating an increasing rate based on the modified risk index; S53: When the ascending rate is greater than the preset ascending reference rate, a difference between the ascending rate and the preset ascending reference rate is calculated and used as an ascending rate difference; S54: generating backup monitoring retrieval information based on the rising rate difference, and calling a preset backup monitoring device for linkage based on the backup monitoring retrieval information; S55: generating early warning feedback information based on the modified risk index and the rising rate; S56: Encrypt the backup monitoring retrieval information and the early warning feedback information and synchronously update them to the multi-level early warning strategy library of the security system terminal.

[0016] In a second aspect, the present invention provides an intelligent security warning system based on AIoT, which adopts the following technical solutions: An intelligent security warning system based on AIoT, including: Acquisition module, used to obtain light detection information, temperature parameters and multimodal data; A memory, configured to store an AIoT-based intelligent security early warning method according to any one of the first aspects; The processor loads and executes the program in the memory.

[0017] In summary, this application includes at least one of the following beneficial technical effects: 1. By dynamically optimizing acquisition equipment settings through real-time collection of light and temperature parameters and combining this with analysis of multimodal data, the accuracy and real-time nature of abnormal behavior identification in complex environments are significantly improved. Furthermore, encrypted transmission of acquired data ensures the security of information transmission while enabling efficient resource scheduling and rapid response. 2. By identifying monitoring blind spots and establishing their spatial coordinates, the system then calculates the required rotation and elevation angle differences based on the spatial coordinates. This system then combines focus parameter corrections to automatically calibrate the monitoring equipment. This automatically eliminates blind spots, enhances the system's adaptability in complex environments, and significantly improves the efficiency of capturing abnormal events and the reliability of early warnings. 3. Based on the initial dynamic model generated from crowd dynamics data, the system compares it with a behavioral baseline model to identify the diffusion rate of abnormal areas. Combined with real-time pedestrian flow, the system dynamically calculates the crowd risk index, enabling a quantitative assessment of potential security threats. By constructing a multi-layered dynamic model, the system presents abnormal distribution and risk evolution trends. This not only optimizes model accuracy based on pedestrian flow in normal conditions, but also accurately locates risk sources, predicts diffusion paths, and assesses impact scope in abnormal scenarios, thereby improving the accuracy of security warnings and response efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a method flow chart of the AIoT-based intelligent security warning method of an embodiment of the present application. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0020] An intelligent security early warning method based on AIoT automatically identifies abnormal behaviors in different environments, reduces missed or misjudgment of security systems, and improves the reliability of security systems.

[0021] Reference Figure 1 The present application discloses an intelligent security warning method based on AIoT, which includes: S1: Real-time collection of light detection information and temperature parameters in the natural environment.

[0022] The natural environment refers to the environment in which people live their daily lives. Acquisition equipment refers to hardware that supports multimodal data acquisition, and includes devices such as image detection devices and sound sensors. Light detection information refers to image information obtained by detecting light on the acquisition device. Light detection information is continuously collected and acquired through light intensity sensors preset on the acquisition device and arranged in a matrix. Light detection information includes light intensity, rate of change, shape characteristics of occluded areas, and movement trajectory. Temperature parameters refer to the value of the ambient temperature, which is obtained through a temperature sensor. The acquisition of light detection information and temperature parameters in the environment provides specific data support for subsequent analysis.

[0023] S2: Generate environmental adjustment information based on the light detection information and temperature parameters and output it to the acquisition device to collect multimodal data.

[0024] Environmental adjustment information refers to instructions for adjusting the acquisition equipment based on environmental conditions, such as adjusting the angle and focus parameters of the image detection device on the acquisition equipment. Multimodal data refers to environmental data that integrates multiple dimensions, including visual and audio data. The specific generation method is described in S21 to S28 below. By adjusting the acquisition equipment in different environments before collecting multimodal data, the integrity and quality of data in complex environments are ensured.

[0025] S3: Generate behavioral information based on multimodal data.

[0026] Behavioral information refers to the dynamics of the target object, such as movement trajectory, speed, and body movements. Different multimodal data corresponds to different behavioral information. Behavioral information is obtained by inputting multimodal data into a pre-set behavioral information database for query. The behavioral information database pre-stores a table comparing different multimodal data and corresponding behavioral information. The behavioral information database is created by technicians summarizing and recording the behaviors reflected by different multimodal data.

[0027] S4: When there is abnormal behavior in the behavior information, early warning information is generated based on the environment adjustment information and the behavior information.

[0028] Abnormal behavior refers to an individual's behavioral patterns that significantly deviate from normal behavioral characteristics. These can include illegal intrusion, robbery, theft, and gang fighting. Warning information refers to an alarm signal that contains information such as the type and specific location of the event. The method for determining the presence of abnormal behavior is described in S61 to S68, S681 to S685, and S71 to S77. When abnormal behavior information is detected, the environmental adjustment information is analyzed. Combined with the specific abnormal characteristics of the behavior information, such as the type, frequency, and severity of the abnormal behavior, a comprehensive assessment is performed using pre-set warning rules and algorithmic models to determine the level and content of the warning. This information is then integrated into a clear, accurate, and targeted warning message. This is promptly sent to the relevant responsible individuals or management departments via pre-set communication channels such as SMS, email, and system push notifications, allowing them to quickly take appropriate countermeasures to ensure safety and stability. Accurately identifying abnormal events and linking them to the environment improves intervention efficiency and enhances the accuracy of warnings. Warning rules and algorithmic models are existing technologies and will not be elaborated on here.

[0029] S5: Trigger a hierarchical warning mechanism based on the warning information, and encrypt and transmit the warning information to the security system terminal for early warning.

[0030] The graded warning mechanism refers to the response level divided according to the severity of the incident, such as level one alarm, level two alarm, level three alarm, etc. Different alarm modes can be set by yourself. Encrypted transmission refers to the end-to-end encryption of data. In this embodiment, the national secret SM4 algorithm is used to encrypt the video stream for transmission, and the key dynamic update cycle is ≤5 minutes. The security system terminal refers to the central platform for receiving and processing alarms. Different warning information triggers different graded warning mechanisms. Different warning information corresponds to different graded warning mechanisms. The graded warning mechanism determines the triggered warning level by querying the preset warning level database. The warning level database pre-stores a comparison table of different warning information and the corresponding warning levels. The warning level database is formed by the operator recording the warning levels corresponding to different warning information in sequence. According to different warning mechanisms, corresponding responses are executed and encrypted transmission is performed to achieve a safe and reliable multi-level emergency response, ensuring that the warning information is tamper-proof and leak-proof.

[0031] Generating environmental adjustment information based on light detection information and temperature parameters includes the following steps: S21: When the light intensity in the light detection information is lower than the preset light reference intensity, the rate of change of light shading, the shape characteristics of the shading area and the movement trajectory within a preset time period are obtained from the light detection information.

[0032] Light intensity refers to the amount of light energy received per unit area, which reflects the strength of the light. Light baseline intensity refers to the pre-set normal light interval value. Preset time period refers to a pre-set time period, such as 30 seconds. The rate of change refers to the speed of change in light intensity caused by the movement of the obstruction. Shape characteristics refer to the geometric properties of the outer contour of the obstruction. Movement trajectory refers to the movement path of the obstruction in the monitoring area. The shape of the obstruction is extracted from the light detection information by using image processing algorithms. The moving speed and path of the obstruction are calculated by target tracking method. Image processing algorithms and target tracking methods are both existing technologies and will not be described in detail here. By obtaining the rate of change, the shape characteristics of the obstruction area and the movement trajectory, the dynamic characteristics of the obstruction are accurately identified to provide data support for the subsequent process.

[0033] S22: Establish a dynamic occlusion model based on the change rate, shape characteristics and movement trajectory.

[0034] A dynamic occlusion model is a mathematical model that describes the motion patterns of obstructions. The rate of change, shape characteristics, and movement trajectory are input into a time series analysis model, which outputs a motion model of the obstruction. This is prior art and will not be detailed here. The model is used to build upon the raw data, making it easier to determine whether the obstruction is caused by human intervention.

[0035] S23: Determine whether the dynamic occlusion model conforms to a preset artificial dynamic occlusion model.

[0036] The artificial dynamic occlusion model is a model that meets the characteristics of artificial occlusion. The artificial dynamic occlusion model is pre-set. Using a similarity comparison algorithm, the dynamic occlusion model is compared with the artificial dynamic occlusion model. If the similarity exceeds the set similarity value, the model is considered to be consistent, thereby reducing false positives and preventing further harm. The similarity comparison algorithm is existing technology and will not be described in detail here.

[0037] S24: If it meets the requirements, the preset UAV is controlled based on the movement trajectory and shape features to perform airspace filling and collect modal filling data. At the same time, the preset collection device is controlled based on the movement trajectory to perform reverse displacement compensation, and the collected data is used as compensation data.

[0038] A drone is a device that can capture images from the air without requiring a human pilot. Aerial positioning refers to filling in monitoring blind spots. A drone's flight path is planned based on the movement and shape characteristics of the obstructed area to ensure coverage. The drone is then controlled to fly along this path, using its onboard image sensor to collect high-altitude bird's-eye view imagery, including environmental images. This data is transmitted to a monitoring center for preprocessing to improve quality. Finally, using spatiotemporal alignment and image fusion techniques, the drone data is integrated with ground-based monitoring data to fill in monitoring blind spots. Modal positioning refers to supplemental high-altitude bird's-eye view imagery collected by the drone. Inverse displacement compensation involves adjusting the angle or position of the image sensor on the acquisition device inversely based on the direction of obstruction to track the target. Compensation data refers to supplemental data, such as visible light and infrared images, location information, and video data collected by the drone. This data is combined with ground-based equipment data to form a complete monitoring dataset. If human occlusion is detected, the drone is commanded to take off to supplement the data in the obstructed area. Through drone and dynamic equipment adjustments, monitoring blind spots caused by human occlusion are eliminated, ensuring data continuity and enabling better tracking of obstructing individuals. Spatiotemporal alignment and image fusion technologies are existing technologies and will not be described in detail here.

[0039] S25: Generate spatiotemporal correlation data based on the modal filling data and the compensation data.

[0040] Spatiotemporal correlation data refers to data generated by time-synchronizing drone high-altitude imagery with the imagery captured by the image detection device on the acquisition equipment. A fusion algorithm is used to generate multi-angle monitoring results based on the information collected by the drone and ground equipment. This fusion algorithm is currently available and will not be described in detail here. By constructing a global monitoring picture, target tracking capabilities are enhanced.

[0041] S26: Generate spatiotemporal adjustment information based on the spatiotemporal correlation data and the temperature parameter, and use the spatiotemporal adjustment information as the environment adjustment information.

[0042] Spatiotemporal adjustment information refers to an adjustment strategy generated based on spatiotemporal correlation data and temperature parameters, used to guide precise adjustments of environmental control equipment. Different spatiotemporal correlation data and temperature parameter combinations correspond to different spatiotemporal adjustment information. This information is retrieved by querying a pre-set spatiotemporal adjustment information database. This database stores a table comparing different spatiotemporal correlation data and temperature parameter combinations with the corresponding spatiotemporal adjustment information. This database is generated by operators testing and measuring the spatiotemporal adjustment information generated by different spatiotemporal correlation data and temperature parameter combinations. By adjusting environmental information, scheduling resources is optimized.

[0043] S27: If not, generate obstacle avoidance information based on the dynamic occlusion model.

[0044] Obstacle avoidance information refers to instructions for avoiding natural obstructions, such as adjusting the angle of the image detection device on the acquisition device to avoid areas with falling leaves. If it does not meet the criteria for human obstruction, it indicates natural obstruction, which can be avoided through active response measures. When the dynamic occlusion model detects a moving target and determines it to be natural occlusion, the coordinate conversion algorithm is used to locate the position parameters of the obstructed area in the field of view of the image detection device on the acquisition device. The angle range is adjusted, and then the horizontal / vertical rotation angles and priority are calculated. A control instruction containing the target avoidance coordinates, rotation direction, and angle is generated. At the same time, the adjusted blind spot coverage rate is verified through edge computing. Finally, a standardized obstacle avoidance protocol is output to drive the device execution, and three sets of data including timestamp, occlusion type, and adjustment parameters are recorded in the log. By actively avoiding natural obstructions, the device's response to non-threatening obstructions is reduced, reducing the false alarm rate.

[0045] S28: Generate environmental fusion information based on the obstacle avoidance information and the temperature parameter, and use the environmental fusion information as environmental adjustment information.

[0046] Environmental fusion information refers to the control information that integrates obstacle avoidance information and temperature parameters. Different combinations of obstacle avoidance information and temperature parameters correspond to different environmental fusion information. This information is retrieved by querying a pre-set environmental fusion information database. This database stores a table comparing different combinations of obstacle avoidance information and temperature parameters with their corresponding environmental fusion information. The database is generated by the operator sequentially recording the environmental fusion information generated for each combination of obstacle avoidance information and temperature parameters. This enables intelligent monitoring that adapts to environmental changes, balancing safety and efficiency.

[0047] Generating environmental adjustment information based on light detection information and temperature parameters also includes the following steps: S291: When the illumination intensity of the illumination detection information is higher than the preset illumination reference intensity, the area with the illumination intensity higher than the preset illumination reference intensity is defined as a highlight area, and at the same time, it is determined whether the temperature parameter is higher than the preset temperature reference parameter.

[0048] Highlight areas are those where the current light intensity exceeds the preset baseline light intensity. The temperature baseline parameter is a pre-set standard temperature value for the current environment, used to determine whether the ambient temperature is too high. By determining whether the temperature parameter under high light conditions is higher than the temperature baseline parameter, the current natural environment is judged to be normal. By identifying areas of excessive light in the current environment and making a preliminary assessment of the temperature situation, a foundation is laid for subsequent environmental adjustments that take both light and temperature factors into account.

[0049] S292: If the temperature is higher than a preset temperature reference parameter, define the area higher than the preset temperature reference parameter as a high temperature area.

[0050] High-temperature zones are areas where the temperature exceeds the reference temperature parameter. Accurately identifying these areas of excessive temperature allows for targeted temperature adjustments to mitigate the effects of high temperatures on equipment.

[0051] S293: Generate a high-light thermal zone based on the high-light zone and the high-temperature zone.

[0052] High-light and high-temperature zones are areas composed of both high-light and high-temperature zones—that is, regions with both high light and high temperatures. Based on the previously defined high-light and high-temperature zones and the preset spatial coordinate system, their ranges and locations are determined. The intersection of these two areas is analyzed and defined as the high-light and high-temperature zone. This identifies areas most affected by light and temperature, facilitating targeted adjustments to environmental parameters.

[0053] S294: Obtain the position of the currently preset shading plate as the current position of the baffle.

[0054] A sunshade is a pre-installed plate used to block light, pre-installed above the image detection device on the acquisition device. The current position of the sunshade refers to the current angular position of the sunshade. This position is determined by reading the angle obtained by the angle sensor pre-installed on the sunshade and analyzing it. This data is then defined as the current position of the sunshade. Understanding the current position of the sunshade provides data support for subsequent adjustments to the sunshade.

[0055] S295: Calculate the rotation angle of the light shielding plate as the adjustment angle based on the current position of the baffle and the high-light hot zone.

[0056] The rotation angle refers to the rotation angle of the sunshade from its current position to the angle at which it can effectively block the high-light and hot areas. According to the current blocking range of the sunshade and the range of the high-light and hot areas, the specific angle to which the sunshade needs to be rotated is determined through mathematical methods of geometric calculation, in preparation for the subsequent adjustment of the sunshade. Using the position and range of the high-light and hot areas and the current preset position of the sunshade, the specific angle to which the sunshade needs to be rotated is determined through mathematical methods such as geometric calculation and trigonometric function operations, so that it can accurately block the high-light and hot areas. For example, in a two-dimensional plane coordinate system, based on the coordinates of the center position of the high-light and hot areas and the current position coordinates of the sunshade, combined with parameters such as the sunshade's shading direction and effective shading range, the angle calculation formula in coordinate geometry is used to calculate the angle to which the sunshade needs to be rotated, and this calculation result is used as the adjustment angle.

[0057] S296: Generate device adjustment information based on the adjustment angle.

[0058] Equipment adjustment information refers to the specific instructions and parameter information required to adjust the sunshade angle. Different adjustment angles correspond to different equipment adjustment information. The equipment adjustment information is obtained by querying a pre-set equipment adjustment information database. The equipment adjustment information database pre-stores a table comparing different adjustment angles with corresponding equipment adjustment information. The equipment adjustment information database is generated by the operator sequentially recording equipment adjustment information generated for different adjustment angles.

[0059] S297: Control the preset shading plate to execute based on the device adjustment information, and simultaneously obtain the illumination adjustment detection information and temperature adjustment parameters after the execution.

[0060] Light adjustment detection information refers to the light intensity after the sunshade performs the shading operation. The light intensity information obtained by the light intensity sensor is used to evaluate the shading effect. For example, after the sunshade is rotated to a certain angle, the light intensity in the area measured by the light intensity sensor drops from 10,000 lux to 500 lux. This 500 lux is the light adjustment detection information. The temperature adjustment parameter refers to the temperature after the sunshade performs the action. The ambient temperature data re-measured by the temperature sensor reflects the impact of shading on temperature. For example, after shading, the temperature measured by the temperature sensor drops from the original 35°C to 29°C. This 29°C is the temperature adjustment parameter. It provides feedback data for further evaluation of the shading effect and environmental control.

[0061] S298: Generate an environment control factor based on the light control detection information and the temperature control parameter, and use the environment control factor as the environment adjustment information.

[0062] Environmental control factors refer to adjustment information used to adjust environmental parameters. Different combinations of light control detection information and temperature control parameters correspond to different environmental control factors. These environmental control factors are determined by querying a pre-set environmental control factor database. The database contains a pre-stored table of corresponding environmental control factors for different combinations of light control detection information and temperature control parameters. The database is created by an operator sequentially recording the environmental control factors generated for different combinations of light control detection information and temperature control parameters.

[0063] S299: If the temperature is not higher than the preset temperature reference parameter, the sound data is retrieved from the multimodal data.

[0064] Sound data refers to ambient sound, captured by sound sensors such as microphones. When the temperature parameter is not higher than the preset temperature reference parameter, sound data is retrieved from the multimodal data set to prepare for subsequent environmental adjustments based on sound and other factors.

[0065] S29A: Optimize image acquisition parameters based on illumination detection information.

[0066] Image acquisition parameters refer to the exposure time, sensitivity, white balance, dynamic range, and other setting parameters of the light intensity sensor that affect image quality. In a security system, the optimized values of the image acquisition parameters are calculated based on real-time light data. In a low-light intensity environment, a parameter combination of extending the exposure time, turning on the infrared fill light, and controlling the digital gain is calculated to reduce noise and increase the brightness of the picture. In strong light scenes, wide dynamic range parameters are calculated to balance the contrast between light and dark areas, and the color temperature parameters are adjusted to prevent overexposure of the picture. In the case of sudden light sources, anti-flicker algorithm parameters are calculated and combined with AI to identify local optimization parameters of key areas to ensure that the monitoring picture achieves the best results in terms of detail clarity and color reproduction accuracy, thereby improving the recognition rate of abnormal behavior. A low-light intensity environment refers to an environment where the light intensity is lower than the preset light reference intensity.

[0067] S29B: Generate audio and video adjustment information based on the image acquisition parameters and the sound data, and use the audio and video adjustment information as environment adjustment information.

[0068] Audio and video adjustment information refers to information used to adjust related equipment after comprehensively considering image acquisition information and sound data. Based on the acquired image acquisition parameters and sound data, the sound characteristics are first analyzed, the energy distribution of abnormal frequency bands and the dynamic change trend of volume are extracted, and the directional enhancement parameters are generated in combination with the sound source localization algorithm. At the same time, the brightness compensation value and exposure correction value in the image parameters are feature-encoded, and the audio and video data streams are aligned through the time series association model. Finally, the sound source directional parameters and the image optical compensation amount are integrated, and the audio and video collaborative control instructions containing spatial enhancement weights and frequency band optimization values are output, realizing the coordinated optimization of audio and video acquisition and processing, ensuring the synchronization and coordination of audio and video during environmental monitoring, improving the overall effect and user experience of the security monitoring system, and enabling monitoring personnel to more accurately judge the environmental conditions through audio and video information. The sound source localization algorithm is an existing technology and will not be elaborated here.

[0069] Generating device adjustment information based on the adjustment angle includes the following steps: S2961: Get the monitoring area.

[0070] The monitoring area refers to the range that the image detection device can currently cover and effectively monitor. By reading the image detection device's parameter settings, including orientation angle and focus, the actual range of the area currently being monitored is determined. This understanding of the monitoring area provides data support for subsequent comparisons with the preset monitoring area to identify potential blind spots.

[0071] S2962: When the monitoring area is inconsistent with the preset monitoring area, the unmonitored area is named as a monitoring blind area.

[0072] The preset monitoring area refers to the pre-set range for image acquisition, determined based on the security system's requirements and scene layout. A blind spot is an area that cannot be monitored. The currently acquired monitoring area is compared with the preset monitoring area in terms of shape, size, and location. Any areas that are not covered by the preset monitoring area are designated as blind spots.

[0073] S2963: Obtaining a coordinate range of the monitoring blind spot as the blind spot coordinates based on a preset spatial coordinate system.

[0074] A spatial coordinate system is a reference coordinate system established to determine the position of an object or area in three-dimensional space. It typically uses a fixed point as the origin and three mutually perpendicular coordinate axes, such as the X, Y, and Z axes, to uniquely identify the position of any point in space. Blind spot coordinates refer to the range covered by the coordinate position of a monitoring blind spot in the spatial coordinate system.

[0075] According to the preset spatial coordinate system, the location information of the monitoring blind spot is converted into the coordinate range under the coordinate system to obtain the blind spot coordinates, which provides accurate geometric position parameters for subsequent monitoring area difference calculation and helps to accurately determine the size and location of the monitoring blind spot.

[0076] S2964: Calculate the monitoring area difference based on the blind spot coordinates.

[0077] The monitoring area difference refers to the difference between the preset monitoring area and the current monitoring area. Based on the blind spot coordinates and the coordinate range of the preset monitoring area, a geometric calculation method is used to determine the area deviation between the two, which is used as the monitoring area difference. This quantitatively reflects the difference between the current monitoring area and the preset target area, providing specific data support for subsequent analysis.

[0078] S2965: Obtain the rotation angle difference and the pitch angle difference based on the monitoring area difference and control the image detection device to execute.

[0079] By adjusting the horizontal rotation angle and vertical pitch angle, the device's observation direction is dynamically adjusted to align with the target location or abnormal area within the monitoring area. The rotation angle difference refers to the angle difference in the horizontal rotation of the image detection device. The pitch angle difference refers to the angle difference in the vertical tilt of the monitoring device. Based on the difference in the monitoring area, trigonometric functions and other calculation methods are used to determine the rotation angle difference and pitch angle difference required for the monitoring device. By determining the specific angle parameters that require adjustment for the monitoring device and instructing the device to perform the corresponding actions, the monitoring area is expanded, blind spots are reduced, and the comprehensiveness and effectiveness of monitoring are improved.

[0080] S2966: Generate a focus parameter correction value based on the rotation angle difference, the pitch angle difference, and the blind spot coordinates.

[0081] The focus parameter correction refers to the focus parameter that enables the device to more clearly image the monitoring blind spot. Different rotation angle differences, pitch angle differences, and blind spot coordinate combinations correspond to different focus parameter corrections. The focus parameter correction is obtained by querying the preset focus parameter correction database to obtain the corresponding correction value. The focus parameter correction database pre-stores a comparison table of different rotation angle differences, pitch angle differences, and blind spot coordinate combinations and the corresponding focus parameter corrections. The focus parameter correction database is formed by the operator conducting sequential test measurements and recording of the focus parameter corrections generated by different rotation angle differences, pitch angle differences, and blind spot coordinate combinations. According to the adjusted posture of the monitoring equipment and the location characteristics of the monitoring blind spot, the focus parameters are accurately adjusted to improve the acquisition quality of the monitoring blind spot image.

[0082] S2967: Generate device adjustment information based on the focus parameter correction amount, the rotation angle difference, and the pitch angle difference.

[0083] Equipment adjustment information refers to the instruction information for adjusting the rotation angle of the image detection device, including parameters such as the rotation angle difference, the pitch angle difference, and the focus parameter correction. The focus parameter correction, rotation angle difference, and pitch angle difference are integrated and generated according to a specific data format and communication protocol to generate complete equipment adjustment information. For example, the rotation angle difference, pitch angle difference, and focus parameter correction are each encoded as a specific instruction code or numerical parameter, combined into a data frame or instruction packet, and sent as equipment adjustment information to the monitoring device's control system. Through comprehensive and precise equipment adjustment instructions, the monitoring device can be guided to adjust its direction and focus parameters simultaneously, achieving effective coverage of monitoring blind spots and high-quality image acquisition.

[0084] The method for determining whether abnormal behavior exists includes the following steps: S61: Extracting behavioral features from multimodal data.

[0085] Behavioral characteristics refer to the significant attributes or key information of a target object's behavioral pattern. Common behavioral characteristics include the target object's motion trajectory, such as speed, direction, and path shape; posture changes, such as the angles of various body parts; movement frequency, such as the number of times a certain action occurs per unit time; and dwell time. By extracting behavioral characteristics, we provide basic data analysis for subsequent behavioral analysis.

[0086] S62: When the behavior characteristics do not conform to the preset behavior benchmark model, extract items that do not conform to the behavior benchmark model from the behavior characteristics and mark them as abnormal items.

[0087] The behavioral baseline model refers to a pre-set behavioral model of normal behavior. Abnormal items refer to feature items in behavioral characteristics that do not conform to the behavioral baseline model.

[0088] S63: Extract the number of abnormal items and use it as the abnormal number.

[0089] The number of anomalies refers to the number of abnormal items in the behavior signature. This provides a simple and intuitive quantitative indicator to measure the degree of behavioral anomaly, providing basic data support for further assessment of the risk level of abnormal behavior and determining whether an alarm is required.

[0090] S64: Generate anomaly confidence based on the abnormal items and abnormal quantities.

[0091] Anomaly confidence refers to the degree of confidence that the current behavior is considered abnormal, typically a value between 0 and 1. A larger value indicates a more likely abnormal behavior; a smaller value indicates a more likely normal behavior. Using a pre-established confidence calculation model, such as a logistic regression model, the anomaly item and number of anomalies are used as input features to calculate the anomaly confidence. Logistic regression models are state-of-the-art and will not be discussed here. This model comprehensively considers the type and number of anomalies to generate a quantitative indicator that more comprehensively reflects the likelihood of an anomaly. This makes subsequent judgments about abnormal behavior more scientific and reasonable, avoiding potential misjudgments based on a single factor.

[0092] S65: Determine whether the abnormality confidence exceeds a preset abnormality reference confidence.

[0093] The abnormal baseline confidence level is a pre-set value used to determine whether behavior is abnormal. By determining whether the behavior exceeds the preset abnormal baseline confidence level, the system determines whether an alert is necessary. This ensures that the system can issue alerts appropriately based on actual security needs and avoid excessive false positives or missed negatives.

[0094] S66: If it exceeds, it is determined that abnormal behavior exists.

[0095] When the judgment result shows that the abnormality confidence exceeds the preset value, the current behavior is directly determined to be abnormal behavior.

[0096] S67: If it does not exceed, the number of times the abnormal item appears is retrieved from the behavior feature as the number of abnormalities.

[0097] The number of anomalies refers to the total number of times an abnormal item occurs. If the anomaly confidence level does not exceed the preset value, further analysis is performed based on the number of anomalies to avoid overlooking potentially recurring abnormal behaviors due to a single, accidental anomaly.

[0098] S68: If the number of anomalies exceeds a preset number, a secondary verification is performed based on the multimodal data.

[0099] The preset number of occurrences refers to a pre-set number of abnormal items, for example, 5, which is used to determine whether further verification is required. Secondary verification involves further analysis and confirmation of the current behavior using comprehensive multimodal data to determine whether abnormal behavior exists. If the number of abnormalities exceeds the preset number, the secondary verification process begins, conducting a more in-depth analysis of the multimodal data. For detailed analysis steps, refer to S681 to S685.

[0100] Secondary verification based on multimodal data includes the following steps: S681: Retrieve sound data from multimodal data.

[0101] Retrieve sound data from multimodal data. Retrieve sound data to provide data support for the subsequent generation of acoustic pattern feature data.

[0102] S682: Generate voice pattern feature data based on the voice data and behavior characteristics.

[0103] Acoustic behavior signature data refers to comprehensive data that fuses acoustic data and behavioral features, used to more comprehensively describe the target subject's behavior. The integration of acoustic data and behavioral features using feature fusion algorithms is a well-established technique and will not be detailed here. By fusing acoustic and behavioral features, it is possible to more accurately identify and distinguish normal from abnormal behavior, improving the accuracy and reliability of behavioral analysis.

[0104] S683: Match the acoustic pattern feature data with the preset acoustic pattern database and extract the one with the highest matching degree as the initial abnormality judgment value.

[0105] The acoustic behavior pattern database stores a large number of pre-set samples of known acoustic behavior pattern feature data, including both normal and abnormal behavior feature data. The initial abnormality judgment value refers to the matching value corresponding to the sample found in the acoustic behavior pattern database that most closely matches the current acoustic behavior pattern feature data. This value indicates the degree of similarity between the current feature data and the known behavior pattern and can be used to preliminarily determine whether the behavior is abnormal. Using a pattern matching algorithm to individually match the current acoustic behavior pattern feature data with all samples in the acoustic behavior pattern database is a conventional technique and will not be discussed here. Using known acoustic behavior pattern feature data as a reference narrows the scope of analysis and reduces the workload for subsequent verification.

[0106] S684: Calculate the difference between the initial abnormality judgment value and the reference matching degree as the abnormality difference.

[0107] The benchmark match is the ideal match standard and is pre-set. The abnormal difference is the difference between the initial abnormality judgment value and the benchmark match, and is used to quantify the degree of deviation between the current behavior and the normal behavior pattern. The abnormal difference, calculated by taking the difference between the initial abnormality judgment value and the benchmark match, provides a quantitative indicator to measure the degree of deviation between the current behavior and the normal behavior pattern, providing a more specific basis for subsequent determination of abnormal behavior.

[0108] S685: When the abnormal difference exceeds the preset abnormal difference interval, it is determined that abnormal behavior exists.

[0109] The abnormal difference interval is a pre-set range used to determine whether the abnormal difference exceeds the normal range. If it exceeds the abnormal difference interval, it is determined to be abnormal behavior, providing a clear judgment result for security warnings and facilitating accurate intervention.

[0110] The method for determining whether abnormal behavior exists further includes the following steps: S71: Obtain current crowd density from multimodal data.

[0111] The current crowd density refers to the number of people within the monitoring area at the current moment. This is calculated by identifying the number of features corresponding to pre-defined person characteristics from multimodal data and then proportionally calculating the number of people to the area corresponding to the pre-defined monitoring area. Person characteristics refer to the characteristics of various body parts and are obtained through pre-entry. The area corresponding to the monitoring area is also obtained through pre-entry.

[0112] Real-time statistics of the number of people in the monitoring area provide basic data support for subsequent crowd density judgment and group behavior analysis, ensuring that the system can understand the distribution of people in a timely manner.

[0113] S72: Determine whether the current crowd density exceeds a preset density value.

[0114] The crowd density threshold is a pre-set threshold based on safety standards, venue capacity, and historical experience. By determining whether the current crowd density exceeds the preset threshold, excessive crowds can be detected, providing a basis for subsequent crowd control measures or enhanced monitoring.

[0115] S73: If the density value is exceeded, the movement trajectory, movement speed and distribution characteristics of the crowd are collected in real time.

[0116] Movement trajectory refers to the path of movement of a group of people or individuals over time. Movement velocity refers to the speed at which a group of people or individuals move. Distribution characteristics refer to the spatial distribution of a group of people, including the degree of evacuation and distribution uniformity. By capturing these parameters using image detection devices, detailed dynamic information about the crowd can be obtained in a timely manner in crowded situations. This provides data support for subsequent group behavior analysis and identification of abnormal individuals, helping to detect potentially dangerous behavior or abnormal dynamics in advance.

[0117] S74: Generate group dynamic data based on the movement trajectory, movement speed and distribution characteristics.

[0118] Crowd dynamics data refers to a data set that comprehensively reflects the overall movement and distribution of a crowd. It includes information such as the crowd's average speed, primary movement direction, changes in clustering areas, and evacuation trends. It describes the dynamic evolution of a crowd over time. The real-time collected movement trajectories, speeds, and distribution characteristics are comprehensively analyzed and processed. First, the average speed and standard deviation of the crowd are calculated to determine the primary movement direction. Next, the temporal changes in clustering areas are analyzed. For example, clustering algorithms are used to detect the displacement of cluster centers and changes in clustering intensity. This is a well-known technique and will not be detailed here. Finally, this information is integrated into a structured database containing fields such as timestamps, average speed, primary direction, clustering area coordinates, and clustering intensity to generate crowd dynamics data. This is achieved through a custom data structure. This generates comprehensive data that reflects the overall movement of the crowd, providing concise and comprehensive data input for the subsequent development of crowd dynamics models, facilitating further analysis and modeling.

[0119] S75: Build a crowd dynamics model based on crowd dynamics data and current crowd density, and mark individual behavior characteristics.

[0120] A crowd dynamics model is used to describe and predict the dynamic behavior of a crowd under various conditions. Individual behavioral characteristics refer to the specific behavioral attributes exhibited by each individual in a crowd, such as sudden changes in walking direction, unusual acceleration or deceleration of speed, and deviation from the mainstream group's direction. For the specific steps for building a crowd dynamics model, refer to S751 to S757. By establishing a crowd dynamics model, the overall behavior of a crowd can be simulated and predicted, while also identifying individual behavioral characteristics. First, a crowd dynamics model is constructed using crowd dynamics data and current crowd density. This model captures the overall dynamic behavior of the crowd. Then, for each individual in the crowd, their behavior is monitored in real time. Based on predefined characteristic criteria, such as sudden changes in walking direction, unusual acceleration or deceleration of speed, and deviation from the mainstream group's direction, the individual is determined to possess these specific abnormal behavioral characteristics. Once an individual is identified as possessing a characteristic, it is recorded and identified using a specific marking method, such as a specific color or symbol in the model visualization interface. This helps accurately identify individuals with abnormal behavior within a crowd, providing strong support for subsequent actions.

[0121] S76: Generate voice data based on individual behavior characteristics and voice data.

[0122] Acoustic data refers to a comprehensive data type that combines individual behavioral characteristics and sound characteristics, used to more comprehensively describe an individual's behavioral state. Labeled individual behavioral characteristic data is fused with sound data. First, the sound data is preprocessed, for example, by removing noise, extracting unique sounds, and identifying specific sound events associated with the individual. This is conventional technology and will not be discussed here. Then, the individual behavioral characteristics are combined with the corresponding sound characteristics to form acoustic data. This can be achieved through data association algorithms, which are conventional technology and will not be discussed here. For example, based on timestamp and spatial location information, individual sound events are matched with behavioral characteristics to generate an acoustic data structure containing fields such as individual identification, behavioral feature vector, and sound feature vector. By fusing individual behavioral and sound characteristics, a richer data description of the individual's state is generated, which helps to more accurately identify abnormal individual behavior, avoid misjudgments caused by relying solely on a single feature, and improve the accuracy and reliability of abnormal behavior detection.

[0123] S77: If the sound behavior data does not match the preset sound behavior reference data within the preset time period, the abnormal individual is marked and determined to have abnormal behavior.

[0124] The preset time period refers to a predetermined time interval, such as 30 seconds before and 1 minute after an event. The acoustic baseline data refers to the acoustic data representing normal behavior and is preset. An abnormal individual is an individual in a group whose acoustic data differs significantly from the acoustic baseline data. Their behavior may pose a threat to safety or require special attention, such as an individual who suddenly walks against the crowd at a high speed and shouts. By comparing the acoustic data obtained during the preset time period and finding that it does not match the preset acoustic baseline data, abnormal individuals can be identified, providing timely abnormal behavior warnings for security monitoring and enhancing the detection capabilities and response speed of abnormal behavior in complex scenarios.

[0125] Building a crowd dynamics model based on crowd dynamics data and current crowd density includes the following steps: S751: Generate an initial dynamic model based on the group dynamic data.

[0126] The initial dynamic model is a preliminary model constructed based on crowd dynamics data, used to describe information such as the average movement speed and primary movement direction of a crowd. This crowd dynamics data is input into a pre-defined model architecture to generate the initial dynamic model. By continuously inputting new crowd dynamics data, the model parameters are dynamically modified using incremental learning, enabling online learning updates. This provides a foundation for further analysis and adjustments, ensuring the system can promptly respond to changes in crowd dynamics. The model architecture can be the continuity equation and the equation of motion from fluid mechanics. Generating models based on this model architecture is a well-established technique and will not be elaborated upon further.

[0127] S752: Determine whether the initial dynamic model is consistent with the group behavior benchmark model based on the initial dynamic model.

[0128] A crowd behavior benchmark model is a pre-established reference model that represents normal crowd behavior patterns. It is typically based on extensive historical data and behavioral characteristics under normal scenarios. By determining whether the initial dynamic model is consistent with the crowd behavior benchmark model, it is possible to effectively identify whether the initial model accurately reflects normal crowd behavior, thereby determining whether further adjustments are needed to improve the model's accuracy and reliability.

[0129] S753: If they match, the initial dynamic model is adjusted based on the current crowd density to form a group dynamic model.

[0130] When the initial dynamic model matches the crowd behavior benchmark model, the current crowd flow is added to the initial dynamic model and parameters are adjusted to obtain a crowd dynamic model. This improves the model's adaptability to the current crowd state, enabling it to more accurately predict and describe crowd behavior, providing more reliable data support for security decision-making.

[0131] S754: If not, extract the abnormal area from the initial dynamic model.

[0132] Abnormal regions are areas in the initial dynamic model that differ significantly from the baseline model. If the initial dynamic model does not match the baseline group behavior model, the abnormal regions need to be extracted from the initial dynamic model. First, a baseline group behavior model must be determined, representing normal crowd behavior patterns. Then, a difference index is calculated between the initial dynamic model and the baseline model. A range is set. When the difference index exceeds the range, an abnormal region is identified. These areas of significant difference in the initial model are then located and extracted as abnormal regions. These extracted abnormal regions are fed into the subsequent risk assessment and response modules to provide a basis for analysis. For example, in a surveillance video scenario, the initial model describes crowd flow, while the baseline model represents normal flow patterns. When the calculated difference index between the two models exceeds the range, the abnormal regions are located and extracted. These may be areas of crowd gathering or abnormal individual movement. These regions can be identified by calculating the difference index between the models. Extracting abnormal regions helps to promptly detect abnormal behavior or distribution within a crowd, providing a basis for subsequent risk assessment and response measures.

[0133] S755: Calculate the area change rate of the abnormal area within a preset time period to generate the diffusion speed of the abnormal area.

[0134] The preset time period refers to a time interval predetermined based on the actual application scenario and is used to monitor changes in the area of the abnormal area. The area change rate refers to the rate of change in the area of the abnormal area within the preset time period. The abnormal area diffusion rate refers to the speed at which the abnormal area expands within the preset time period, and is obtained based on the area change corresponding to the abnormal area during the preset time period. The ratio of the area change to time is then calculated to obtain the area change rate, and the abnormal area diffusion rate is deduced based on the area change rate. Different area change rates correspond to different abnormal area diffusion rates. The abnormal area diffusion rate is obtained by querying a preset abnormal diffusion database to obtain the corresponding value. The abnormal area diffusion database pre-stores a comparison table of different area changes and the corresponding abnormal area diffusion rates. The abnormal diffusion database is formed by scientific researchers through continuous dynamic monitoring of changes in the abnormal area diffusion rate caused by different area changes. It provides data support for taking measures in advance and helps prevent potential safety risks.

[0135] S756: Calculate the group risk index based on the diffusion rate of the abnormal area and the current crowd density.

[0136] The group risk index is a comprehensive indicator reflecting the risk level of the current crowd situation, taking into account the spread rate of the anomaly zone and the current crowd density. Data on the spread rate of the anomaly zone and the current crowd density are collected. A risk assessment model uses these two parameters as input variables to calculate the group risk index, helping to quickly identify high-risk scenarios and implement appropriate security measures. Risk assessment models are existing technology and will not be discussed in detail here.

[0137] S757: Establish a multi-layer group model based on the initial dynamic model, abnormal areas and risk index, and use the multi-layer group model as the group dynamic model.

[0138] A multi-layered group model is a comprehensive model consisting of multiple layers. The model's anomaly layer utilizes information from abnormal regions to represent areas of abnormal behavior or distribution. Risk indices are combined to create a risk layer, reflecting the degree of risk in different areas. These layers are combined with the initial dynamic model or other relevant data to form a multi-layered model structure. Ultimately, the multi-layered group model, as a group dynamic model, can more comprehensively and accurately describe and predict the dynamic behavior and risk profile of a population.

[0139] After encrypting and transmitting the warning information to the security system terminal for early warning, the following steps are also included: S51: Generate a modified risk index based on the environmental adjustment information and multimodal data.

[0140] The modified risk index is the result of modifying the original risk index, taking into account environmental adjustments and multimodal data to more accurately reflect the risk level of the current environment and behavior. A risk assessment model uses environmental adjustments and multimodal data features as input variables to calculate the modified risk index. This comprehensively considers the impact of environmental and behavioral factors on risk. Input variables include adjusted light intensity and temperature, as well as crowd density, movement speed, and the number of unusual behaviors from multimodal data. The model output is the modified risk index, which reflects the comprehensive risk level of the current environment and population status.

[0141] S52: Generate an increase rate based on the modified risk index.

[0142] The rising rate refers to the speed at which the modified risk index changes within a unit of time, and is used to measure the changing trend of the risk level. The rate of change of the modified risk index within a unit of time is calculated and used as the rising rate. For example, if the unit of time is 15 seconds, the modified risk index at each time point within this 15 seconds is recorded, and then the index difference between adjacent time points is calculated, and the average rate of change is calculated, which is the rising rate. Calculating the rising rate provides a quantitative indicator to measure the changing trend of the risk level, which helps to detect the rapid increase of risk in a timely manner and provide a basis for subsequent emergency measures. S53: When the ascending rate is greater than the preset ascending reference rate, the difference between the ascending rate and the preset ascending reference rate is calculated and used as the ascending rate difference.

[0143] The baseline rate of ascent is a pre-set threshold used to determine whether the rate of ascent of the modified risk index is abnormal. The rate of ascent differential measures the degree of abnormality of the current ascent rate. The current ascent rate is compared with the preset baseline rate of ascent. If the current ascent rate is greater than the preset value, the difference between the two is calculated and stored as the rate of ascent differential, providing a basis for subsequent activation of backup monitoring equipment.

[0144] S54: Generate backup monitoring retrieval information based on the rising rate difference, and call the preset backup monitoring equipment for linkage based on the backup monitoring retrieval information.

[0145] Backup monitoring retrieval information refers to the retrieval instructions and parameters for backup monitoring devices. Backup monitoring devices are pre-configured backup monitoring devices used to provide additional monitoring capabilities when primary monitoring devices are insufficient or require enhanced monitoring. The urgency of the backup monitoring retrieval is determined based on the magnitude of the rise rate difference. Based on the urgency, devices to be retrieved are screened from the pre-set backup monitoring device list and used as backup monitoring retrieval information. The backup monitoring device list details the identification, location, monitoring range, device type, corresponding retrieval priority, and applicable scenarios of all backup devices. The backup monitoring retrieval information is linked to the pre-set backup monitoring devices to improve monitoring accuracy.

[0146] S55: Generate early warning feedback information based on the modified risk index and the rising rate.

[0147] Early warning feedback information refers to comprehensive information including the revised risk index and escalation rate, providing security personnel with detailed feedback on the current risk situation. Different revised risk index and escalation rate combinations correspond to different early warning feedback messages. Early warning feedback information is obtained by querying a pre-set early warning feedback information database. The early warning feedback information database pre-stores a table comparing different revised risk index and escalation rate combinations with corresponding early warning feedback messages. The early warning feedback information database is generated by operators sequentially recording early warning feedback messages generated for different revised risk index and escalation rate combinations. This helps security personnel quickly understand the current risk situation and supports decision-making.

[0148] S56: Encrypt the backup monitoring retrieval information and the early warning feedback information and synchronously update them to the multi-level early warning strategy library of the security system terminal.

[0149] The multi-level early warning strategy library is a database of multi-level early warning strategies stored in the security system terminal, used to implement appropriate early warning measures based on different risk levels. Backup monitoring access information and early warning feedback are encrypted, using the national SM4 algorithm for encrypted video stream transmission. The key is dynamically updated every 5 minutes or less, ensuring data transmission security. The encrypted data is then synchronized and updated to the multi-level early warning strategy library in the security system terminal, improving the reliability and responsiveness of the security system.

[0150] Based on the same inventive concept, an embodiment of the present invention provides an AIoT-based intelligent security warning system, including: Acquisition module, used to obtain light detection information, temperature parameters and multimodal data; A memory for storing programs such as the aforementioned AIoT-based intelligent security warning method program; The processor loads and executes the program in the memory.

[0151] Those skilled in the art will clearly understand that for the sake of convenience and brevity, the division of the above-mentioned functional modules is only used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working processes of the above-mentioned systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0152] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that improvements and modifications that do not depart from the principles of the present invention are within the scope of protection of the present invention.

Claims

1. An intelligent security early warning method based on AIoT, characterized in that: include: S1: Real-time collection of light detection information and temperature parameters in the natural environment; S2: generating environmental adjustment information based on the light detection information and the temperature parameter and outputting the environmental adjustment information to a collection device to collect multimodal data; S3: generating behavioral information based on the multimodal data; S4: When the behavior information contains abnormal behavior, generating warning information based on the environment adjustment information and the behavior information; S5: triggering a hierarchical warning mechanism based on the warning information, and encrypting and transmitting the warning information to the security system terminal for warning.

2. The intelligent security early warning method based on AIoT according to claim 1, characterized in that: Generating the environment adjustment information based on the light detection information and the temperature parameter includes: S21: When the light intensity in the light detection information is lower than a preset light reference intensity, obtaining a change rate of light shading, shape characteristics of the shading area, and a movement trajectory within a preset time period from the light detection information; S22: establishing a dynamic occlusion model based on the change rate, the shape feature, and the movement trajectory; S23: Determine whether the dynamic occlusion model conforms to a preset artificial dynamic occlusion model; S24: If the conditions are met, a preset drone is controlled based on the movement trajectory and the shape feature to perform airspace filling and collect modal filling data, and a preset collection device is controlled based on the movement trajectory to perform reverse displacement compensation, and the collected data is used as compensation data; S25: generating spatiotemporal correlation data based on the modal filling data and the compensation data; S26: generating spatiotemporal adjustment information based on the spatiotemporal correlation data and the temperature parameter, and using the spatiotemporal adjustment information as the environment adjustment information; S27: If not, generating obstacle avoidance information based on the dynamic occlusion model; S28: Generate environmental fusion information based on the obstacle avoidance information and the temperature parameter, and use the environmental fusion information as the environmental adjustment information.

3. The AIoT-based intelligent security early warning method according to claim 2, characterized in that: Generating the environment adjustment information based on the light detection information and the temperature parameter further includes: S291: When the illumination intensity of the illumination detection information is higher than a preset illumination reference intensity, defining a region higher than the preset illumination reference intensity as a highlight region, and determining whether the temperature parameter is higher than a preset temperature reference parameter. S292: If the temperature is higher than a preset temperature reference parameter, define the area higher than the preset temperature reference parameter as a high temperature area; S293: generating a high-light-heat zone based on the high-light zone and the high-temperature zone; S294: Obtain the current preset position of the light shield as the current position of the baffle; S295: Calculating a rotation angle of the light shielding plate as an adjustment angle based on the current position of the baffle and the high-light hot zone; S296: Generate device adjustment information based on the adjustment angle; S297: Controlling the preset shading plate to execute based on the device adjustment information, and simultaneously obtaining illumination adjustment detection information and temperature adjustment parameters after execution; S298: Generate an environment control factor based on the light control detection information and the temperature control parameter, and use the environment control factor as the environment adjustment information; S299: If the temperature is not higher than the preset temperature reference parameter, retrieving the sound data from the multimodal data; S29A: Optimizing image acquisition parameters based on the illumination detection information; S29B: Generate audio and video adjustment information based on the image acquisition parameters and the sound data, and use the audio and video adjustment information as the environment adjustment information.

4. The AIoT-based intelligent security early warning method according to claim 3 is characterized in that: Generating device adjustment information based on the adjustment angle includes: S2961: Get monitoring area; S2962: When the monitoring area is inconsistent with the preset monitoring area, the unmonitored area is named as a monitoring blind area; S2963: Obtaining a coordinate range of the monitoring blind spot as the blind spot coordinates based on a preset spatial coordinate system; S2964: Calculating a monitoring area difference based on the blind spot coordinates; S2965: Calculating a rotation angle difference and a pitch angle difference based on the monitoring area difference and controlling the light intensity sensor to execute; S2966: Generate a focus parameter correction value based on the rotation angle difference, the pitch angle difference, and the blind spot coordinates; S2967: Generate the device adjustment information based on the focus parameter correction amount, the rotation angle difference, and the pitch angle difference.

5. The intelligent security early warning method based on AIoT according to claim 1, characterized in that: Methods for determining abnormal behavior include: S61: extracting behavioral features from the multimodal data; S62: When the behavior feature does not conform to the preset behavior benchmark model, extracting items that do not conform to the behavior benchmark model from the behavior feature and marking them as abnormal items; S63: extracting the number of abnormal items and using it as the abnormal number; S64: generating anomaly confidence based on the abnormal item and the abnormal quantity; S65: Determine whether the abnormality confidence exceeds a preset abnormality reference confidence; S66: If it exceeds, it is determined that the abnormal behavior exists; S67: If not, the number of occurrences of the abnormal item is retrieved from the behavior feature as the number of abnormalities; S68: If the number of abnormalities exceeds a preset number, a secondary verification is performed based on the multimodal data.

6. The AIoT-based intelligent security early warning method according to claim 5, characterized in that: The secondary verification based on the multimodal data includes: S681: Retrieving sound data from the multimodal data; S682: Generating voice behavior feature data based on the voice data and the behavior feature; S683: Matching the acoustic pattern feature data with a preset acoustic pattern database and extracting the one with the highest matching degree as an initial abnormality judgment value; S684: Calculate the difference between the initial abnormality judgment value and the reference matching degree as the abnormality difference; S685: When the abnormal difference exceeds a preset abnormal difference interval, it is determined that the abnormal behavior exists.

7. The AIoT-based intelligent security early warning method according to claim 6, characterized in that: Methods for determining abnormal behavior also include: S71: Obtaining current crowd density from the multimodal data; S72: Determine whether the current crowd density exceeds a preset density value; S73: If the density value is exceeded, the movement trajectory, movement speed and distribution characteristics of the crowd are collected in real time; S74: generating group dynamic data based on the movement trajectory, the movement speed, and the distribution characteristics; S75: Establishing a group dynamics model based on the group dynamics data and the current crowd density, and marking individual behavior characteristics; S76: generating sound behavior data based on the individual behavior characteristics and the sound data; S77: If the sound behavior data does not match the preset sound behavior reference data within the preset time period, the abnormal individual is marked and determined to have the abnormal behavior.

8. The AIoT-based intelligent security early warning method according to claim 7, characterized in that: Establishing a group dynamics model based on the group dynamics data and the current crowd density includes: S751: generating an initial dynamic model based on the group dynamic data; S752: Determine whether the initial dynamic model is consistent with the group behavior benchmark model based on the initial dynamic model; S753: If they match, adjusting the initial dynamic model based on the current crowd density to form a group dynamic model; S754: If not, extracting an abnormal area from the initial dynamic model; S755: Calculating the area change rate of the abnormal region within a preset time period to generate the diffusion speed of the abnormal region; S756: Calculating a group risk index based on the diffusion speed of the abnormal area and the current crowd density; S757: Establishing a multi-layer group model based on the initial dynamic model, the abnormal area, and the risk index, and using the multi-layer group model as the group dynamic model.

9. The AIoT-based intelligent security early warning method according to claim 8, characterized in that: After encrypting and transmitting the warning information to the security system terminal for warning, the method further includes: S51: generating a modified risk index based on the environmental adjustment information and the multimodal data; S52: generating an increasing rate based on the modified risk index; S53: When the ascending rate is greater than the preset ascending reference rate, a difference between the ascending rate and the preset ascending reference rate is calculated and used as an ascending rate difference; S54: generating backup monitoring retrieval information based on the rising rate difference, and calling a preset backup monitoring device for linkage based on the backup monitoring retrieval information; S55: generating early warning feedback information based on the modified risk index and the rising rate; S56: Encrypt the backup monitoring retrieval information and the early warning feedback information and synchronously update them to the multi-level early warning strategy library of the security system terminal.

10. An intelligent security warning system based on AIoT, characterized in that: include: Acquisition module, used to obtain light detection information, temperature parameters and multimodal data; A memory, configured to store an AIoT-based intelligent security warning method according to any one of claims 1 to 9; The processor loads and executes the program in the memory.

Images